I am not sure where you have read the Noctua D15 hit 95C with 3950x or 3900x with stock frequency. I have a 3900x with Noctua D15 and I am getting 76C with 100% cpu utilization with ~165W package power. It is a fact: not myth! 80C@~200W package power. I had made it very clear in the past post. All this is for stock and undervoltage without overclocking/PBO. Wraith Prism works because it is measured against package power with default locked power limit. (~165W)
Let me be clear, I never said that it did. I said that air cooling wasn't recommended, nor sufficient to achieve the maximum potential performance of a 3950X. It seems that people running on air cooling get lower clocks than those who aren't. There may be something else in the boost clock algorithm that causes this when specific temperatures are reached. That is, if your temp hits X amount in X time at 100% utilization, it downclocks to a lower speed. I do not know. AMD doesn't make its boost clock algorithm details known. However, AMD is clear in its press kit for the 3900X that better cooling will produce better results. They even used the D15 as an example of good air cooling. Again, AMD doesn't recommend air cooling at all for the 3950X.
Was this test done while cpu was at 100% utilization? When my PBO is on auto so I see that all core can go 4250 before but drop out like 5 sec later to 4100/4050.
Of course. Mine starts off at 4.3GHz or higher and then drops down. It settles at 4.12GHz using AutoOC in Cinebench R20 and remains there until the test completes. Frankly, Cinebench R20 is a bad test for this as it's over too quickly.
All,
I am not sure you guys understand what I had been implying but I will make it explicit. Please post your package power along with your temperature. Under utilized cpu will run cooler than fully utilized cpu at same frequency or higher frequency. Prime95/OCCT does not fully max out cpu utilization. Cinebench R20 can peak fully at one point but does not constantly. PBO is consider to be overclocking up to +200MHz on top of PB2. PB2 is turbo boosting: it is not overclocking and is within AMD specification.
I did. 40% under PB2 and 41% under PBO or auto-overclocking. Under PB2, temperature is 68c. Under PBO/AutoOC its 79c or more at full load using Blender / Gooseberry. Under Cinebench R20, it only goes to about 76c because the test doesn't run long enough to allow the CPU to get warmer. And yes, I understand that PBO is overclocking. PB2 is stock operation, but again, even so your clocks are influenced by the same values. If you run weaker cooling, you may be limited by your CPU temps before you run into other limits. It looks like AMD targets 68c for these under PB2 and as a result, the use of different cooling solutions will produce different results. A better cooler will allow for higher clock speeds. AMD even states this is the case.
I stand by my statement. A D15 will probably work, but air cooling isn't recommended and it isn't enough in my opinion. If you use air coolers with the 3950X, you will lose performance. Whether or not that's acceptable to you is another matter entirely.
And I do not want to install that shitty Wraith Prism on the 3950X just to prove that it won't be worth a shit.